Financiación del BEI para infraestructuras de transporte

Type of relationship n=19 n=62 n=47

Have family relations 46.4 52.7 53.7 Member of the same club 4.8 3.1 0.0 Live in the same village 98.9 95.6 97.3 Live in the same hamlet (para) 91.0 80.3 85.3 Member of another NGO group as well 1.6 8.1 7.5

Source: IFPRI 1994.

Table 3.5 Fluctuation of members, by type of group Program

RDRS BRAC ASA

n=19 n=62 n=47

Members now 17.1 36.9 18.1

Members five years ago 19.7 56.2 25.0 Members two years ago 16.6 49.8 19.9

Members quitting 2.3 17.8 4.7 Members joining 1.43 6.3 3.4 Fluctuation indexa Mean 10.5 42.4 32.2 Coefficient of variation 152.5 55.4 89.9 Source: IFPRI 1994.

a_{Fluctuation index =[(number of members joining +number of members quitting) /}

acres in the case of ASA and BRAC mem- bers. This is consistent with conclusions reached by a number of other studies. Mor- duch (1998b), using the data set of the study by Pitt and Khandker (1998), shows that 28 percent of Grameen Bank members and 21 percent of BRAC members had initial land- holdings above 0.5 acres of land at the time of joining the program. Zaman (1997) finds that 28 percent of borrowers from BRAC are above the eligibility criteria.

Obviously, given their small landholding, most members derive a significant part of their income from nonagricultural sources. The percentage reporting farming as the main occupation is significantly higher for RDRS

than for the other groups. This may be due to the fact that RDRS, supported by the Marginal and Small Farms Systems Crop Intensification Project (MSFSCIP), takes in members who have as much as 1.5 acres of land, whereas ASA and BRAC limit eligibil- ity to households owning less than 0.5 acres. The fact that a large proportion of mem- bers—with the exception of RDRS—derive most of their income from nonagricultural sources has important implications for the financial institutions: because of the wider spectrum of activities in the nonagricultural sector, the incomes of group members are likely to be less correlated. This may have a positive impact on repayment performance, Table 3.6 Characteristics of group members, by NGO program

Program

RDRS BRAC ASA

n=325 n=2,315 n=880

Age of members (years) 35.2 33.3 32.0

Member is head of household (%) 67.1 22.8 13.8

Share of women (%) 28.6 86.4 98.0

Household size of member 4.8 4.8 4.8

Number of household members under 14 years of age 1.9 1.8 2.0 Cultivable land owned by household (decimala_{) 47.4 41.1 49.7}

Occurrence of major sickness/death (%)b _{15.7 18.3 20.9}

Occurrence of crop loss (%)b _{17.2 28.4 21.6}

Occurrence of major social event (%)b _{17.8 21.7 27.4}

Level of education (%)

Illiterate 76.3 79.7 81.0

Primary education 11.1 10.0 11.4

Secondary education 9.8 8.9 7.3

Higher education 2.8 1.4 0.3

Major occupation of member (%)

Household work 27.1 64.4 66.4 Farmer 14.8 3.1 4.9 Large business 0.0 0.0 0.0 Small business 9.8 12.9 19.8 Salaried professional 2.2 2.3 0.2 Day laborer 36.6 8.2 2.8 Craftsman 1.2 3.0 4.4 Fisherman 0.6 0.8 0.0 Rickshaw puller 3.4 0.7 0.0 House servant 2.2 2.7 0.8 Other 2.1 1.6 0.7 Source: IFPRI 1994. a_{100 decimals =1 acre.} b_{Over previous 18 months.}

since members can potentially bail each other out of household-specific income shocks. The Conduct of Groups

The conduct of groups is influenced both by regulations stipulated by the NGOs as well as by regulations that are internally de- cided upon by the members. Table 3.7 not only presents the major functions of the groups, but also indicates how priorities

have changed over time. Respondents were asked to rank functions in order of priority at two points in time: during the year of group formation and during the survey year. Most of the functions listed in Table 3.7 are exter- nally stipulated.

Clearly, saving was the most important function during the first year of group formation in all three credit programs. This is because all three credit programs require Table 3.7 Ranking of functions of groups in the first year and in survey year (%)

Program

RDRS BRAC ASA All

n=19 n=62 n=47 n=128

Rank Rank Rank Rank Rank Rank Rank Rank Rank Rank Rank Rank

1 2 3 1 2 3 1 2 3 1 2 3

Saving together

In the first year 100 0 0 100 0 0 89 11 0 96 4 0

In 1994 90 13 0 55 44 0 53 47 0 59 41 0

Receiving credit

In the first year 0 0 0 0 27 39 11 52 39 4 35 36

In 1994 11 44 0 45 55 0 47 53 0 41 53 0

Purchasing inputs or selling outputs together

In the first year 0 0 0 0 0 2 0 0 0 0 0 1

In 1994 0 0 0 0 0 2 0 0 0 0 0 1

Investing jointly in business

In the first year 0 50 0 0 0 4 0 0 0 0 4 2

In 1994 0 31 46 0 0 0 0 0 4 0 4 8

Receiving training in business management, production technologies, and marketing

In the first year 0 20 57 0 12 14 0 2 0 0 9 13

In 1994 0 6 31 0 0 14 0 0 30 0 1 21

Receiving training in gender issues, human rights, and social awareness

In the first year 0 0 43 0 7 14 0 2 19 0 4 14

In 1994 0 0 8 0 2 10 0 0 19 0 1 12

Improving education for children

In the first year 0 0 0 0 0 0 0 0 0 0 0 0

In 1994 0 0 0 0 0 53 0 0 0 0 0 30

Improving education for adults

In the first year 0 20 0 0 54 22 0 30 35 0 42 27

In 1994 0 0 8 0 0 8 0 0 33 0 0 15

Solving social problems

In the first year 0 0 0 0 0 2 0 0 4 0 0 2

In 1994 0 0 0 0 0 6 0 0 0 0 0 3

Helping each other in crisis

In the first year 0 0 0 0 0 2 0 0 4 0 0 2

In 1994 0 0 8 0 0 0 0 0 7 0 0 3

Other functions

In the first year 0 10 0 0 0 2 0 2 0 0 2 1

In 1994 0 6 0 0 0 11 0 0 7 0 1 7

members to save initially. RDRS, for ex- ample, requires members to save during the training period, which takes place before the groups are certified as creditworthy. BRAC and ASA have similar conditionalities. There was a significant drop in the number of groups in ASA and BRAC that still regarded saving as the most important activity at the time of the survey in 1994. In the case of RDRS, although saving activities continued to be the main priority in 1994, groups saved only a minimum with the banks. Whenever savings increased beyond the minimum stipulated, they were withdrawn. This may indicate that saving is mainly the result of conditionalities imposed on loans by finan- cial institutions and that households view it merely as part of the cost of obtaining a loan. It could also indicate that the interest rates offered by the financial institutions on sav- ings are lower than yields on investment elsewhere.

Groups with RDRS do not borrow at all in the first year of formation, and no BRAC groups reported borrowing to be the most important activity. In contrast, 11 percent of ASA groups ranked receiving credit first. The relatively few credit transactions in the year of group formation indicate considerable screening and monitoring of group members prior to the initiation of a full-fledged lend- ing program. This time may also be used by the members themselves to assess each others’ creditworthiness. Training and adult education are important functions in the ini- tial year of group formation, but their role generally declines over the years.

For the survey year, most groups indi- cated that both credit and saving were the most important services. Close to one-half of the groups belonging to BRAC and ASA ranked both credit and saving as most im- portant. In the case of RDRS, saving con- tinued to be the most important service; only 11 percent ranked credit first. How- ever, even in the case of RDRS, 44 percent reported credit to be the second most im- portant activity.

The Performance of Credit Groups: An Examination of Loan Default Rates

The performance of groups can be evaluated in two general areas: (1) the extent to which they are able to take advantage of the serv- ices provided by the financial institutions and use them to enhance their own welfare, and (2) their performance in complying with the contractual arrangements. The first area will be discussed in Chapters 4 and 5. This sec- tion will focus on the second area, and more specifically on the repayment performance of groups—in other words, the default rate of credit groups. Loan default is one of the most important determinants of banks’costs. Thus, the financial sustainability of NGO- based financial systems, like that of banks, rests on keeping the default rate below 10 if not below 5 percent.

Table 3.8 shows the historical repayment performance of all ASA, BRAC, and RDRS groups combined. In total, the 128 groups (and their respective subgroups of approxi- mately 5 members each) received 1,725 loans of which 876 were due before the date of survey. The group leaders were asked about the repayment status, and the infor- mation provided was verified with the local NGO branch office. Overall, 85 percent of loans were reported to be fully repaid at the due date, but 15 percent were in arrears. The repayment rates are favorable when

Table 3.8 The number of group loans and their repayment rate

Number Percent

Total number of loans obtained 1,725 Due after date of survey 758 Repayment status not clear 91 Due before date of survey 876 Loans for which due date has passed

Fully paid at due date 743 85

Partially paid at due date 133 15

Totally unpaid 0 0

compared with those for government-owned commercial banks, which, in 1992, averaged only 19 percent.

Why Are the Repayment Rates of Group-Based Organizations So Good?

Fairly recent work in institutional economics has shed considerable light on why new group-based institutions have been able to perform so well, whereas others fail.

In group lending programs, the functions of screening, monitoring, and enforcement of repayment are, to a large extent, transferred from the bank’s agent to the borrowers—the group members themselves. It is argued that groups accomplish these tasks better than banks and therefore achieve higher repay- ment rates. Stiglitz (1990) and Varian (1990) discuss these perceived advantages of col- lective action in the screening of loan ap- plicants and monitoring of borrowers. The incentives for screening and monitoring the actions of peers arise from joint liability and the potential loss of access to future loans. The main argument is that, compared with socially and physically distant bank agents, group members can obtain information, at a low cost, on the reputation, indebtedness, and wealth of loan applicants, and about their efforts to ensure the repayment of a loan. Zeller (1994) shows that members of formal groups—like informal lenders—consider a peer’s indebtedness in the informal market as a major determinant of credit rationing. Thus, group members are able to access complex and sensitive information, just like informal lenders. Groups may also have a comparative advantage in the enforcement of loan repayment. Whereas the formal lender usually has limited means to compel repay- ment from delinquent borrowers, group members have the potential to employ social sanctions or seize physical collateral (Besley and Coate 1995). In many rural societies, including those in Bangladesh, commercial bank agents have little leverage actually to go to a village and seize a defaulter’s collat-

eral. Furthermore, group members appear to be in a better position to assess the reason for default and to offer insurance services to those members experiencing shocks beyond their control, while imposing sanctions on willful defaulters.

It is important to note, however, that group lending may not ensure higher repay- ment rates at all times. First, because the risk of loan default by an individual is shared by their peers, a member may choose a riskier project than if it were an individual contract. This may occur because the individual bor- rower counts on other members to repay the loan so that they can secure future loans for themselves. Bratton (1986) analyzes the re- payment record of credit groups in Zimbabwe and shows that expectations about peers’ probability of repaying a loan influence the repayment behavior of an individual member: group loans performed better than individual loans in years of good harvest, but worse in drought years. Varian (1990) argues that such domino-like effects may be mitigated if group members are able to exclude potentially bad borrowers. Similar reasoning underlies the suggestion by Stiglitz (1990) and by Deve- reux and Fishe (1993) that individuals facing a similar magnitude of risks have an incen- tive to form groups.

However, there is also the problem of covariate shocks when the impaired repay- ment ability of some members coincides with the equally impaired capacity of other mem- bers to bail them out. The empirical analysis by Zeller (1998) suggests that individuals may attempt to exploit economies of risks by grouping with others whose income streams are negatively correlated with theirs. In other words, heterogeneity among members with respect to economic activities or risk expo- sure is potentially beneficial for repayment rates. The role of mutual intra-group insur- ance in credit groups is also confirmed by Sadoulet and Carpenter (1999), who find that risk heterogeneity among group mem- bers in Guatemala facilitates mutual help. The sustainability of group lending programs

in areas with high covariate risks depends on the ability of the financial intermediary to reschedule defaulting members’ loans or to raise funds from borrowers during a normal year to cover such contingencies.

Lastly, there is the question of the optimal group size. Groups beyond a certain size may experience increased difficulty in exchang- ing information and in coordination. Further, the disincentives to reneging on contracts diminish, because each member may expect the effect of their action on other members to be diluted (Glance and Huberman 1994).

To sum up, although the empirical evi- dence suggests that the repayment records of group-based credit systems are much better than those of traditional commercial banks, economic theory still suggests situ- ations where groups may actually perform poorly. From a policy point of view, it is important to know more about these types of situation, so that changes can be made in institutional design to minimize their impact.

Econometric Analysis of Default Rates

The dependent variable used in this study is the default rate (DEFAULT), defined as the percentage of debt in arrears at the date when complete repayment was promised. DE- FAULT=0 implies the complete repayment is on time, whereas DEFAULT=100 implies complete default. There were no cases of the latter.

The default function is defined as follows: DEFAULT=f(LNAMNT,X, Z, M), (3.1) where LNAMNTis the loan size, Xis a vec- tor of group characteristics, Zis a vector of community characteristics, and Mis a vector of lender characteristics. Note that this function is defined only for LNAMNT>0. A function is specified with the property that Lim_LNAMNT→0 DEFAULT=0.This is a reasonable assumption, since defaults on small amounts of loans are indeed likely to

be zero. When equation (3.1) is a linear function, this specification is achieved by in- teracting X, Z, Mwith LNAMNT,as in equa- tion (3.2). A corollary of this assumption is that the effects of X, Z,and Mon the default rate are made conditional on the loan size, that is,

∂(DEFAULT)

———————=g(LNAMNT), ∂X

and similarly for Zand M.

Also, because the dependent variable is truncated at zero (the group decides not to default), the estimating equation is specified more generally as (for the i’th group)

DEFAULT_i* = β₁(LNAMNT) +(LNAMNT)Xβ₂+(LNAMNT)Zβ₃ +(LNAMNT)Mβ₄+e_i, (3.2) where DEFAULT_i=0 if DEFAULT_i* ≥0 and DEFAULT_i=DEFAULT_i* if DEFAULT_i* >0.

In this framework, DEFAULT_i* is a latent variable observable only when it takes a positive value. Equation (3.2) is estimated by using the Tobit maximum likelihood technique (Maddala 1983), after correcting for heteroskedasticity, based on the method proposed by Greene (1993). This model was implemented on a subset of the data set that included only those transactions whose due date had passed at the time of the inter- view (n=876 in Table 3.8) and for which information on X, Z,and Mwas completely available. These transactions totaled 868 loans given to the subgroups of all 128 groups.

Regressors, Hypotheses, and Discussion of Results

Table 3.9 presents the results of the Tobit maximum likelihood estimation of the de- fault equation. GROUPSIZE is defined as the average size of a subgroup in a group. If there are no subgroups, GROUPSIZE is equal to the size of the group. The hypothesis is that, the bigger the group, the more likely it is that information flows are imperfect between members. Hence, problems arising out of asymmetric information make monitoring and enforcement costly and less effective. Rates of default are therefore expected to increase with group size. The sign of the coefficient is positive as expected; however, it is insignificant at the 10 percent level.

LNAMNT and (LNAMNT)2 _{are the}

value of the loan, in taka, and its square, re- spectively. Two factors are at work. First, the greater the loan size, the greater the proba-

bility of default. Second, the larger the loan, the higher is the borrower’s cost of delaying payment [=(1 +r+p)* LNAMNT], where p is the incremental penalty rate of interest. The second factor puts pressure on the borrower to reduce late payments. Consideration of this factor is important, because default in the sample appears to mostly consist of ar- rears that are eventually paid, even if they are paid late (as opposed to complete default). A squared term is included for this reason. The coefficient on LNAMNT is positive and significant and therefore supports the first part of the hypothesis. Though the sign of the coefficient on the squared term is nega- tive, as expected, it is not significant.

M_LAND is the mean amount of land owned by the group. Since it reflects owner- ship of an important asset, it was expected that it would enhance the capacity of the group to repay loans on time. In the equation, the effect of landownership on the default rate is found to be negative and significant, as expected. This indicates the importance of even a marginaldifference in the amount of land owned, since all three programs, es- pecially BRAC and ASA, limit their lending to persons in households owning less than 0.5 acres of land. This result may be partly due to the high marginal productivity of land at such low levels.

VARLAND is the variance of the land owned by members of a particular group. This variable was used as one indicator of the portfolio diversity among members of a group. It was hypothesized that, the greater the diversity, the less the covariance of the incomes. Hence, a higher variance was ex- pected to be associated with a lower rate of default, because it would enable a better pool- ing of risk among members. A similar meas- ure of portfolio diversity was used by Zeller (1998) in relation to the repayment perform- ance of credit groups in Madagascar. The coefficient is negative, but not significantly different from zero. The insignificance may also be due to the fact that both ASA and BRAC use a strict criterion for landowner-

Table 3.9 Determinants of default on group loans (Tobit)

Variablea _{Mean Unit Coefficient t−ratio}

LNAMNT 12.031 taka 0.11 × 10−4 _4.922** (LNAMNT)2 _25.5×107 _{−0.35 × 10}−11 _−0.23 GROUPSIZE 12.5 number 0.18 × 10−7 _1.48 M_LAND 0.50 acres −0.14 × 10−7 _−2.06** VARLAND 1.62 −0.33 × 10−6 _−0.73 RATION 25.0 percent −0.54 × 10−7 _−3.85** (RATION)2 _{5,140.0 0.46 × 10}−10 _2.26** RELATIVES 51.5 percent 0.19 × 10−7 _1.82* SHOCKS 22.0 −0.46 × 10−7 _−2.68** AG_PROP 0.3 percent −0.56 × 10−5 _−2.88** M_DRT 0.35 percent −0.19 × 10−4 _−4.43** PCFEMALE 87.0 percent −0.57 × 10−7 _−6.73** DUMINTD 0.30 0.15 × 10−5 _3.60** LN_AGE 1.55 years −0.35 × 10−7 _−0.15 DISTANCE 12.0 miles −0.18 × 10−6 _−2.19** SAMITY 0.23 number 0.97 × 10−6 _1.612* FFW 0.23 dummy variable −0.11 × 10−5 _−1.63* IRRI 30.0 percent 0.18 × 10−7 _1.88* PARTRATE 200.0 per í000 − 0.69 × 10−8 _−3.86** DUMRDRS 0.013 dummy −0.18 × 10−5 _−0.26 DUMBRAC 0.71 dummy 0.41 × 10−5 _2.71** Log likelihood = −438.27

*=significant at 10 percent level; **=significant at 5 percent level.

ship of 0.5 acres or less as one of their eligibility requirements.

RATION is computed as the difference between the value of the loan applied for and the actual value of the loan received, expressed as a percentage of the total loan amount. A higher degree of rationing im- plies a higher level of unfulfilled credit de- mand. If this generates a greater concern for future borrowing privileges, groups can be expected to increase efforts to lower de- fault rates. However, if the degree of ra- tioning is too high, it is likely to render the

In document Metodología para la creación de un Fondo de Garantía de Infraestructuras y evaluación de su eficiencia en su aplicación a concesiones de infraestructuras de transportes (página 108-113)